Research On Online Grinding Chatter On Monitoring Methods For CNC Grinding Machine

Grinding is a high-precision processing technology, but in the grinding process a forcedexcitation frequency is close to a weak natural frequency of the grinding system, which willlead to the grinder resonance occurs and seriously affect the quality of work piece. Thevibration signal in grinding process is a typical non-stationary and nonlinear signal, so theamount of extracted features that can reflect the grinder fault state information from vibrationsignals has become the most important problem to solve the grinding quality.Conventional vibrating signal processing is based on Fourier, but Fourier transform is aglobal transform and lack of physical meaning for the analysis of non-stationary signals.Response to these problems, this thesis is supported by the Zhejiang Province Natural ScienceFund Projec（tLZ13E050003）.Based on researching the vibration of the grinding processing, thenew method-HHT is introduced to analysis the non-stationary vibration signals and extracted thegrinding chatter feature. Least square support vector machine (LSSVM) is seen as recognitionclassifier to indentify the relationship between chatter feature and grinding states, whichdevelops a chatter online detection methods and provids theoretical basis for the development ofengineering applications.The main contents and conclusions are as follows:Firstly, this thesis describes research status and trends of the grinding chatter, and points outthe deficiency and research trend of grinding chatter in the current methods. Hilbert HuangTransform (HHT) methods is proposed to analysis the signals from the grinder processing andextract the features changed significantly from the smooth vibration phase transition to chatter inthe process of the grinder, which can used to identify the correspondence relationshipbetween the chatter feature and grinding chatter;Secondly, according to the feature of the vibration signals in the grinding processing, theappropriate sensors and signal acquisition system are chose to build the platform for grinderchatter test. By changing the grinding parameters, the signals of different mode are collected tosupply the data for subsequent research;Finally, LSSVM is introduced into the grinding chatter for grinding chatter judgment andrecognition. Real-time variance and instantaneous energy which were extracted based on HHTare used to training and learning with LSSVM, diagnostic model is constructed for grinding chatter to judge and identify the grinding chatter. Vibration signals collected by grinding test isused to verify the grinding chatter identification method based on HHT and LSSVM can be usedas the basis for judgment and identification grinder chatter occurs.